用共轭梯度法改善AR谱估计性能
IMPROVEMENT OF PERFORMANCE OF AR SPECTRAL ESTIMATION BY THE CONJUGATE GRADIENT METHOD
-
摘要: 本文基于AR模型前向预测误差最小,提出用共轭梯度法搜索AR模型参数,从而实现谱估计。计算机仿真表明,用这种方法进行谱估计,不会产生谱漂移现象,谱分裂得到明显减弱,而且还具有很高的频率分辨能力。用这种方法估计的信号频谱性能接近于Marple算法。
-
关键词:
- 谱估计; 共轭梯度法; 谱漂移; 谱分裂
Abstract: A new spectral estimation method based on the least mean square error of the forward linear prediction of AR model is introduced in this paper. The spectral estimation is obtained by the AR parameters estimated by the conjugate gradient method. The simulation results show that for this algorithm there is no bias in the frequency estimation, and the spectral line splitting is very week and the spectral resolution is high. The performance of spectral estimation. of the new method is similar to that of the Marple algorithm. -
L. Marple, IEEE Trans. on ASSP, ASSP-28 (1980) 4, 441-454.[2]关肇直等,信号分析与处理,科学出版社,北京,1986年,第275-337页.[3]刘德贵等,FORTRAN算法汇编(第二分册),国防工业出版社,北京,1988年,第471-1473页.[4]黄俊钦,随机信号处理,北京航天航空大学出版社,北京,1990年,第68-102页.[5]黄俊钦等,一种ARMA谱估计公式,自动化学报,12(1986)10,33-40.[6]J. A. Cadzow, IEEE Trans. on ASSP, ASSP-26(1978)5,524-528. -
计量
- 文章访问数: 2354
- HTML全文浏览量: 154
- PDF下载量: 531
- 被引次数: 0
下载:
